Ceph Squid on Proxmox VE 9: Plan Your Upgrade Before the Deadline

Ceph Squid (19.2) is the default Ceph release in Proxmox VE 9, and it quietly sets a calendar deadline for every cluster still on Reef: Squid’s patch window closes on 19 September 2026. If you are running Reef, the upgrade is not a distant project — it is this quarter’s work.

Why the deadline matters

Ceph releases have a fixed support horizon, and Proxmox tracks them closely. Reef (18.2) is now in the rear-view mirror for new installs, and running past a release’s patch window means operating without security and bug fixes for the storage layer underneath every virtual machine. The good news is that the Reef-to-Squid upgrade path is well documented and, for a healthy cluster, genuinely uneventful.

The upgrade itself

Ceph upgrades are rolling and online: each monitor, manager and OSD is restarted one at a time while the cluster keeps serving data. Set the noout flag so the cluster does not rebalance while you work, upgrade monitors first, then OSDs, and only move to the next node once the cluster is healthy again. Patience is the whole technique — a single step too early is how people turn a routine upgrade into a recovery.

Size and network it properly

Squid does not change Ceph’s fundamentals. You still need a dedicated, ideally 10 GbE-or-faster network for Ceph traffic, separate from management and guest traffic, and you still store everything three times with the default three replicas. Capacity planning has not changed: budget for the replication factor and for headroom, because a full OSD is an emergency, not a warning.

When Ceph is the right call

Ceph earns its complexity when you need horizontal scale, shared storage for live migration, and no dependency on external storage hardware — typically a half-dozen nodes or more, each contributing disks. For a small three-node cluster, a well-built ZFS pool with replication is often the more maintainable answer. Choose the storage, then the release, not the other way round.